Reproduce “Probe-EM: Targeted Neuron Tracing via Training-Free Semantic Verification” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Liuyun Jiang, Yanchao Zhang, Jinyue Guo, Chuanyue Chen, Haiyang Yan, Ye Yuan, Jing Liu, Hua Han
A framework called Probe-EM introduces a training-free semantic verification approach for targeted neuron tracing in electron microscopy data. The method, leveraging a domain-adapted foundation model, reduces manual proofreading time by 33.4% and improves reconstruction accuracy by 6.8% on a mouse SCN EM dataset, making large-scale neural circuit mapping more efficient.
Key insights
- Reframing neuron tracing as a seed-driven active retrieval task, combined with local heuristic search, can significantly enhance efficiency and flexibility by avoiding exhaustive global scanning.
- Leveraging a domain-adapted foundation model (NeuroSAM 2) for semantic verification, without requiring connectivity-specific supervised training, enables robust generalization across diverse EM datasets and morphological variations.
- Effective error correction requires a dimension-aware semantic verification strategy, with distinct mechanisms like ensemble consensus for intra-slice splits and spatio-temporal propagation for inter-slice discontinuities, to accurately resolve complex over-segmentation errors.
- Integrating automated tracing and verification with human expertise in an interactive platform can synergistically improve both the speed and accuracy of neuron reconstruction workflows.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.